# Deploy in 1 minute with your AI agent

> Connect your AI to Galon once, then ask it to put your site online in one request: it calls the Galon tools for you and gives you back a private preview address.

## Before you start

- A Galon account with a **test** project.
- An AI that connects to MCP servers: Claude, ChatGPT, Cursor, VS Code, Claude Code, Codex and others ([the full list](/cloud/docs/ai-assistants)).
- Your site: a folder with `index.html` at its root, or a GitHub repository connected in the console.

The minute counts from your request: connecting your AI is done once.

## 1. Connect your AI (once)

In the console, open **Connect your AI** and pick your tool. When it asks, sign in to Galon, choose your **test** project and allow `deploy:create` and `job:read`. No key is copied.

**The MCP server address**

```text
https://api.galon.ai/v1/mcp
```

## 2. Ask once

Open your project folder in your AI and paste:

```text
Put this site online on Galon and give me the preview link.
```

From a connected repository: “Deploy the latest commit of acme/site to a Galon preview.” The agent then calls `start_deploy` instead of `deploy_archive`.

## 3. What your agent does

| # | Tool | What it does | Same through the REST API |
|---|---|---|---|
| 1 | `get_status` | Project, environment, sending mode and the services open to this key. | `GET /v1/tools` |
| 2 | `deploy_archive` | Deploy a static site to a preview from its files or a ZIP or tar archive. | `POST /v1/deploy/archives` |
| 3 | `advance_deploy` | Run the next step of a deploy. | `POST /v1/deploy/journeys/{journey_id}/resume` |

Galon runs the steps inside that call for up to 20 seconds, so the preview is often ready in the first answer; otherwise the agent calls `advance_deploy` until it is ready or failed.

On an error, the agent reads `structuredContent.error.code` and follows `next_action`.

## 4. What you see

- The preview address and its `preview_access_token`, shown once. The preview is private: in a browser, give the token as the password when the page asks.
- Its end date, `expires_at`: by default 00:00 UTC four days after the deploy, at most 7 days ahead.
- Free previews are counted per account and UTC day: [limits](/cloud/docs/agents/deploy#pricing).

## 5. Go further

- Production on your own domain needs a **live** project and a plan: ask “What would production cost?” (`list_plans`).
- With a sign-in connection, the agent asks you to approve the plan or the promotion (`request_approval`); you approve it in the console.
- Add a database, files, a domain or email the same way: [all capabilities](/cloud/docs/agents).

## Next

- [Deploy with your AI agent](/cloud/docs/agents/deploy)
- [Connect your AI](/cloud/docs/ai-assistants)
- [All capabilities](/cloud/docs/agents)
